Re: [PATCH rc v2] iommu/arm-smmu-v3: Fix inconsistent ATS state tracking

[Jason Gunthorpe <jgg@ziepe.ca>]

On Tue, May 05, 2026 at 10:32:40PM +0000, Pranjal Shrivastava wrote:
> The PF shows no ATS Cap:

Yeah, that's complete broken and unusable. We should be failing things
not trying to patch around it. quirks should be used to deal with
this non-compliant HW.

I like the idea of prepare ats failing on the VF if the PF isn't setup
right and then either failing probe or just disabling ATS on that
device.

I also like the idea of failing attach when ats enable fails so we
keep a consistency, but it should be a WARN_ON path that shouldn't
happen do the prepare_ats dealing with it.

> which means pci_ats_supported returns true for such broken HW. Ideally,
> it should return false for VF if the PF doesn't support ATS!

That's also a possible option, but I also wonder if ats_supported and
prepare_ats should maybe be merged together? Why does the iommu driver
have to call both? If prepare succeeds then allow ATS otherwise HW
does not support ATS, turn it off ?

Jason